Assessing Impacts of Climate Change and Human Activities on the Abnormal Correlation Between Actual Evaporation and Atmospheric

Sustainable Cities and Society 56 (2020) 102075

Contents lists available at ScienceDirect

Sustainable Cities and Society

journal homepage: www.elsevier.com/locate/scs

Assessing impacts of climate change and human activities on the abnormal correlation between actual evaporation and atmospheric evaporation T demands in southeastern China

Hua Baia,b,c, Xianghui Lub, Xiaoxiao Yangd, Jianchu Huangd, Xingmin Mua,c,*, Guangju Zhaoa,c, Faliang Guib, Chao Yuea,c a Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources, Yangling, Shaanxi 712100, China b Jiangxi Key Laboratory of Hydrology-Water Resources and Water Environment, Nanchang Institute of Technology, Nanchang, Jiangxi 330099, China c University of Chinese Academy of Sciences, Beijing 100049, China d Jiangxi Provincial Bureau of Hydrology, Nanchang, Jiangxi 330002, China

ARTICLE INFO ABSTRACT

Keywords: Generally, atmospheric evaporation demand (AED) shows a positive correlation with actual evaporation (AE)in Climate change the Yangtze River basin. In order to explore whether and why abnormal correlation exits in its five sub-basins, Afforestation the temporal changes between AED and AE were compared, and then the impacts of climate change and human Penman equation activities on abnormal correlations were assessed using the sensitivity method of Penman equation and the Budyko-type equation decomposition method of Budyko equation. The results indicated: (1) Pan evaporation and potential evaporation Water-energy balance − − (indicators of AED) decreased at rates of up to -0.04 and -0.06 mm d 1 decade 1, respectively. In contrast, the Complementary relationship of evaporation − − water budget-derived evaporation (an indicator of AE) increased at the rate of 0.09 mm d 1 decade 1 in the Fuhe River basin and remained stable in other basins. Abnormal correlations (nonpositive correlations) were observed. (2) Decreasing net radiation and wind speed were major climate factors resulting in simultaneous temporal decreases in AED and AE. In contrast, afforestation was a major human factor leading to an increase in annual AE of 78 mm, but had no effect on AED.Afforestation was the primary driving force of the abnormal correlation. These results could provide water-energy guidance for urban climate change mitigation and flood/ drought disaster management.

1. Introduction correlation (AC) indicates the existence of a regional water-energy balance transformation. Identifying its driving force will be beneficial The mechanism of the water-energy balance transformation is a for understanding the response mechanism of how the water-energy basic, crucially investigated issue in urban hydrology in response to a balance transforms in response to a changing environment. Such an changing environment that is mainly dominated by rapid urbanization, understanding can also specify the variation in the boundary conditions ecological restoration, and climate change (Remondi, Burlando, & in order to predict urban water-energy processes (Ali, Shafiee, & Vollmer, 2016). Characterization of the controlling factors and their Berglund, 2017), which can then be applied to address the urban heat linkages with other environmental elements to form a full picture of the island effect (Giridharan & Emmanuel, 2018; Roberge & Sushama, water-energy cycle has provided a more complete description of the 2018), greenhouse gas emissions (Hardin et al., 2017), drought (Gober, response mechanisms involved (Brutsaert & Parlange, 1998; Yang et al., Sampson, Quay, White, & Chow, 2016), and flood disasters (Nachshon, 2007). Evaporation is the controlling environmental factor because it Netzer, & Livshitz, 2016) in a changing environment. connects the water and energy cycles (Milly, 1994; Wolock & McCabe, Water budget-derived evaporation (WBDE) using the surface water 1999; Yang et al., 2007). The correlation between atmospheric eva- balance method is the proper basin-wide AE index, while potential poration demand (AED) and actual evaporation (AE) quantifies the evaporation (ETp) and pan evaporation (Epan) are the appropriate in- regional water-energy balance. The appearance of an abnormal dicators of AED. Current observational and estimation methods of AE

⁎ Corresponding author at: State Key Laboratory of Soil Erosion and Dryland Farming on the Loess Plateau, Northwest A&F University, 26 Xinong Road, Yangling, 712100, Shaanxi Province, China. E-mail address: [email protected] (X. Mu). https://doi.org/10.1016/j.scs.2020.102075 Received 30 January 2019; Received in revised form 2 September 2019; Accepted 26 January 2020 Available online 31 January 2020 2210-6707/ © 2020 Elsevier Ltd. All rights reserved. H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075 mainly include the lysimeter, eddy covariance, Bowen ratio (Wang & complementary relationship (BCR) between AED and AE, denoted by Dickinson, 2012; World Meteorological Organization (WMO) (2014)), their negative correlation, has been observed in parts of typically en- scintillometer, surface water balance, and atmospheric water balance ergy-limited regions (Wang et al., 2011), especially in the regions methods (Wang & Dickinson, 2012). These methods have different covered with dense forest (Carmona, Poveda, Sivapalan, Vallejo-Bernal, spatial scales. Only the surface and atmospheric water balance methods & Bustamante, 2016; Lawrimore & Peterson, 2000) and urban areas are suitable for the basin scale (up to thousands of kilometres), while (Nakamichi & Moroizumi, 2015). In these regions, the expected positive other methods are applicable for scales below hundreds of meters correlation between AE and AED for the entire energy-limited region (DeBruin, 2009; Moene, Beyrich, & Hartogensis, 2009; Solignac et al., was transformed into a negative relationship. Potential driving forces of 2009; Wang & Dickinson, 2012). However, substantial errors have been this abnormal correlation (AC) were climate change or human activities found in estimates made by the atmospheric water balance method (Arora, 2002). For instance, the aridification of parts of an energy- (Roads, 2003; Wang & Dickinson, 2012). Therefore, the surface water limited region could increase CAI more than the threshold CAI value of balance method is the appropriate method for directly obtaining AE at a water-limited region, thus leading to the AC (Zhang et al., 2012; the basin scale. AED is estimated by reference evapotranspiration (ETr) McVicar, Roderick, Donohue, Niel et al., 2012). Human activities (e.g., or ETp and is measured by Epan. But these two measures are intended for ecological restoration, rapid urbanization) could increase water and different applications. ETr has been applied for irrigation scheduling energy consumption (Hong et al., 2019; Wang, Feng, Zuo, & using prescribed crop coefficients (Allen, Pereira, Raes, & Smith, 1998). Rameezdeen, 2019) to alter the water-energy balance, and eventually

In contrast, Epan and ETp are used for estimating open water evaporation change the positive correlation. Therefore, further investigations are (Fu, Liu, Chen, & Hong, 2004; Fu, Charles, & Yu, 2009; Penman, 1948), needed to determine which variable is driving this abnormal correlation as well as for estimating basin-wide AE according to hypotheses re- between AE and AED: climate change or human activities? This paper garding relationships with AE (Bouchet, 1963; Budyko, 1974; Penman, focuses on river basins in an energy-limited region of southeast China.

1948; Sun, 2007; Yang, Sun, Liu, Cong, & Lei, 2006). Accordingly, Epan These basins have experienced strong afforestation and urbanization in and ETp are the suitable indicators for estimating basin-wide AE. last 30 years compared with adjacent basins. The objective of this study The Bouchet's complementary hypothesis (BCH) and the Penman was to identify the appearance of AC and to determine whether climate hypothesis (PH) are proposed to describe the correlation between AE change, urbanization, or afforestation is the primary driving force of and AED in arid and humid regions, respectively. Specifically, in arid AC. and humid regions, also known as water-limited and energy-limited regions, evaporation is limited by available water and energy, respec- 2. Study areas tively. An intermediate region is limited by both available water and energy. These regions are distinguished by different climatological ar- The Poyang Lake basin is a typical energy-limited region with idity index (CAI) values, defined as the ratio of mean annual ETp to CAI≤0.76. It is a sub-basin of the Yangtze River basin. It was selected precipitation (P). Water-limited and energy-limited regions are defined as the study area to identify and interpret AC. Five tributaries of the with CAI≥1.35 and CAI≤0.76, respectively, while an intermediate Poyang Lake basin were selected, including the Ganjiang, Fuhe, region is characterized by 0.76 < CAI < 1.35 (McVicar, Roderick, Xinjiang, Raohe, and Xiushui Rivers (Fig. 1). These tributaries flow Donohue, & Niel, 2012; Zhang et al., 2012). In energy-limited regions, through southeastern China, with the drainage areas ranging from AED is generally positively correlated with AE according to PH 5,013 to 80,948 km2 and the annual runoff coefficients ranging from (Penman, 1948; Yang et al., 2006). In contrast, in water-limited regions, 0.44 to 0.64 (Table 1). AED is negatively correlated with AE according to BCH (Bouchet, 1963; Zuo et al., 2016). In intermediate regions, the correlation between AE 3. Datasets and methods and AED tends to be explained by both PH and BCH. Even though PH and BCH appear as alternative hypotheses, they can be unified in the 3.1. Datasets Budyko framework (Budyko, 1974; Sun, 2007). Assessment of the temporal trends of AED and AE have been sum- The digital elevation model (DEM) with a 30 m spatial resolution marized to verify these hypotheses. Annual pan evaporation has stea- was acquired from the geospatial data cloud (Computer Network dily decreased since 1960 in most regions across the world, such as Information Center, 2009). Based on this data, the drainage areas of the Russia, Siberia (Golubev et al., 2001; Peterson, Golubev, & Groisman, five study basins were extracted by the hydrology tool of ArcGIS soft- 1995), England and Scotland (Stanhill & Möller, 2008), Italy (Moonen, ware (version 9.3, manufactured by ESRI, Inc.). The land use map with Ercoli, Mariotti, & Masoni, 2002), arid regions of China, eastern China, a 100 m spatial resolution was provided by the Institute of Remote all of China (Liu, Xu, Henderson, & Gong, 2004; Shen, Liu, Liu, Zeng, & Sensing and Digital Earth, Chinese Academy of Sciences for 1985, 1995, Tian, 2010; Xu, Pan, Gao, Fu, & Chiang, 2016; Yang & Yang, 2012), 2000, 2005, and 2010. Annual basin-wide forest coverage rate (FCR) India (Chattopadhyay & Hulme, 1997), Australia, and New Zealand data were collected from the Jiangxi Statistical Yearbook (Bureau of

(Roderick & Farquhar, 2004; 2005). During the same period, annual ETp Statistics of Jiangxi, 2012). The monthly normalized difference vege- has decreased in northwest and southeast China (Chen, Gao, Xu, Guo, & tation index (NDVI) data were obtained from the Ecological Forecasting Ren, 2005; Thomas, 2000), all of China (Wang et al., 2017; Zhang et al., Lab at the NASA Ames Research Center (Pinzon & Tucker, 2014; 2016). 2019), and India (Chattopadhyay & Hulme, 1997), but increased in Basin-wide NDVI values were area-weighted averaged to obtain the Australia and New Zealand (Hobbins, Dai, Roderick, & Farquhar, 2008), value in each basin. The monthly streamflow data from 1961 to 2015 south Florida (Abtew, Obeysekera, & Iricanin, 2010), North America were provided by the Jiangxi Provincial Bureau of Hydrology. (Hember, Coops, & Spittlehouse, 2017), northeast and southwest China The daily meteorological data at each station shown in Fig. 1 were (Chen et al., 2005; Thomas, 2000), and Romania (Croitoru, Piticar, obtained from the China Meteorological Data Sharing Service System. Dragotă, & Burada, 2013). In contrast, the temporal trends of WBDE Data quality was controlled with code provided by the National have been reported as both positive and negative (Ukkola & Prentice, Meteorological Information Center (2019). Stations were excluded 2013), and these trends are overwhelmingly influenced by P. Pre- when the rate of missing data exceeded 20 %. The meteorological cipitation can explain 45–46 % of AE trends in energy-limited basins variables included maximum and minimum air temperature (oC), wind − and 80–84 % in water-limited basins (Ukkola & Prentice, 2013). speed at 10 m height (m s 1), relative humidity (RH, %), sunshine -2 -1 By comparing the temporal trends of AED and AE in the studies duration (hours), precipitation (mm), solar radiation (Rs,MJm day ), -1 mentioned above, the BCH and PH have been confirmed in water-lim- and Epan (mm day ) from 1961 to 2015 (Fig. 1). Mean daily tempera- ited and energy-limited regions, respectively. However, Bouchet's ture (T) was the mean value of daily maximum and minimum air

2 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Fig. 1. Study area in southeastrn China including the Ganjiang (GRB), Fuhe (FRB), Xinjiang (XRB), Raohe (RRB), and Xiushui (XSB) River basins, reservoirs, hydro- meteorological stations, and computational grid points. temperature. Wind speed at 10 m height was adjusted to the wind speed spatial interval of the adjacent points in the grid was chosen to be at 2 m height (u2)(Allen et al., 1998). Among the meteorological sta- 10 km. The computational grid points (Fig. 1) were generated from the tions shown in Fig. 1, observed Rs was only available in three stations. starting point according to the spatial interval using the Meshgrid tool The procedure to calculate annual evaporation-related meteor- in Matlab R2018b manufactured by MathWorks, Inc, and were clipped ological variables, Epan, ETp, and WBDE at the basin scale was as fol- to the study basins. lows: (2) The estimation of annual evaporation-related meteorological vari- (1) The gridded study basins ables at meteorological stations

In each study basin, the starting point of the computational grid Annual evaporation-related meteorological variables were calcu- points was located at the lower left corner of the DEM ASCII ﬁle. The lated using daily data at each meteorological station. Observed Rs

3 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Table 1 A–P model parameters (Allen et al., 1998; Angstrom, 1924; Ozturk, Basic characteristics of five river basins in southeastern China. 2015) as: No. Name Area (km2) Mean annual value (mm) RC n Rs = ⎛ab+ ⎞Ra ⎝ N ⎠ (1) P R WBDE ETp R =+=−RR(1 αRR ) + 1 GRB 80,948 1644 846 798 1015 0.51 nnsnl snl (2) 2 FRB 15,811 1777 782 995 1009 0.44 where a and a+b are the fraction of extraterrestrial radiation reaching 3 XRB 15,535 1873 1167 706 1039 0.62 4 RRB 5013 1657 905 752 1010 0.55 the earth on overcast days and clear days, respectively, N is the max- 5 XSB 3548 1526 982 544 984 0.64 imum possible sunshine duration (hour), Ra is the extraterrestrial ra- −2 -1 diation (MJ m day ), Rns is the net solar or shortwave radiation (MJ Note: R and RC represent mean annual runoff depth and runoff coefficient, -2 -1 -2 - m day ), and Rnl is the net outgoing longwave radiation (MJ m day respectively; P is precipitation; WBDE is water-budget derived evaporation; ETp 1). is potential evaporation; GRB, FRB, XRB, RRB, and XSB are the Ganjiang, Fuhe, We then estimated annual ETp, radiation, and aerodynamic terms of Xinjiang, Raohe, and Xiushui River basins, respectively. the Penman equation (ETpr and ETpa)(Donohue, McVicar, & Roderick, 2010; Penman, 1948; Shuttleworth, 1993) as: values were only available at 23 stations over the region defined by 105−125 °E and 20−35 °N (Table 2). Those stations are named here- ETETETpprp=+a (3) after as solar radiation stations. The A–P Model was employed to esti- Δ mate Rs at meteorological stations where Rs observations did not exist. ETpr = RRLHTRn = ()n Δγ+ Model parameters were calibrated by the ordinary least squares method (4) (OLS) based on the observed Rs and sunshine duration data at each of γ 6430VPD ETpa = (1+= 0.536uETVPDT22 )pa ( , )(1 + 0.536 u ) the solar radiation stations (Angstrom, 1924; Ozturk, 2015). The model Δγ+ λ estimates of R were evaluated by the Nash-Sutcliffeefficiency coeffi- s (5) cient (NSE)(Nash & Sutcliffe, 1970), coefficient of determination (R2), and Student’s t-test (significance level = 0.05) (Gosset, 1942). γ 6430VPD 6430VPD ETVPDTpa (,)= =−[1RLH ( T )] Δγ+ λ λ (6) (3) The estimation of annual evaporation-related meteorological variables, E and ET at grid points RH pan p VPD=−= esa e ⎛1 − ⎞es ⎝ 100 ⎠ (7) − Each type of annual data at meteorological or solar radiation sta- where Δ is the slope of saturation vapour curve (Pa K 1), γ is the − tions was interpolated to the grid points using Kriging or Voronoi psychrometric constant (Pa K 1), RLH is the ratio of latent heat to R , ’ n methods according to the data s probability distribution type (Sarangi, VPD is the vapor pressure deficit (Pa), λ is the latent heat of vapor- Madramootoo, & Enright, 2006). The Kriging method was applicable to −1 ization of water (2.45 MJ Kg ), es is the saturation vapor pressure the data obeying a normal distribution. In contrast, the Voronoi method (kPa), and ea is the actual vapor pressure (kPa). does not require a normal data distribution. The elevations of grid points were extracted from the DEM data by ArcGIS 9.3 software. (4) The estimation of annual evaporation-related meteorological vari- We selected 0.23 as the albedo or canopy reflection coefficient (α) ables, Epan, ETp, and WBDE in the study basins (Allen et al., 1998). At each grid point, the annual Rs and net radiation (Rn) were calculated based on gridded sunshine duration (n) and the The gridded annual meteorological variables, Epan, and ETp were

Table 2 A-P Model parameters and evaluation at solar radiation stations in southeastern China.

2 No. Name LON (°) LAT (°) abNs R Student’s t test value t0.05 NSE

1 Zhongshan 113.4 22.5 0.54 0.15 9248 0.77 174 1.96 0.76 2 Guangzhou 113.3 23.1 0.53 0.16 19694 0.82 296 1.96 0.83 3 Shaoguan 113.6 24.8 0.57 0.15 10744 0.82 223 1.96 0.82 4 Ganxian 114.8 25.8 0.55 0.15 19503 0.88 373 1.96 0.88 5 Fuzhou 119.3 26.1 0.59 0.15 19853 0.86 345 1.96 0.86 6 Changning 112.4 26.4 0.60 0.15 8595 0.77 170 1.96 0.77 7 Jianou 118.3 27.1 0.59 0.19 8526 0.79 178 1.96 0.79 8 Changsha 112.9 28.2 0.55 0.14 10485 0.86 256 1.96 0.86 9 Mapoling 113.1 28.2 0.59 0.15 9115 0.86 237 1.96 0.85 10 Jishou 109.7 28.3 0.56 0.14 8690 0.83 205 1.96 0.82 11 Hongjia 121.4 28.6 0.57 0.18 8708 0.87 239 1.96 0.88 12 Nanchang 116.0 28.7 0.57 0.15 19205 0.87 357 1.96 0.86 13 Lushan 116.0 29.6 0.68 0.15 10131 0.84 232 1.96 0.83 14 Tunxi 118.3 29.7 0.59 0.15 8668 0.88 252 1.96 0.89 15 Cixi 121.2 30.3 0.60 0.13 10488 0.88 281 1.96 0.87 16 Hangzhou 120.2 30.3 0.58 0.14 19509 0.85 335 1.96 0.85 17 Wuhan 114.3 30.6 0.53 0.16 18286 0.83 300 1.96 0.81 18 Yichang 111.1 30.7 0.56 0.15 19673 0.83 315 1.96 0.82 19 Baoshan 121.5 31.4 0.55 0.19 8810 0.84 215 1.96 0.86 20 Hefei 117.3 31.9 0.56 0.15 19036 0.85 326 1.96 0.85 21 Nanjing 118.8 32.0 0.57 0.15 19850 0.87 369 1.96 0.87 22 Lvsi 121.6 32.1 0.55 0.16 8751 0.88 248 1.96 0.88 23 Huaian 119.0 33.6 0.54 0.19 5478 0.81 154 1.96 0.84

2 Note: LON denotes the longitude; LAT denotes the latitude; a and b denote the parameters of A–P model; Ns denotes the number of samples; R denotes the coefficient of determination; t0.05 denotes the critical value of Student’s t-test value; NSE denotes Nash-Sutcliffeefficiency coefficient.

4 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075 area-weighted averaged to obtain the values in each basin. The WBDE of climate change and human activities on AC in order to identify the was estimated as the diﬀerence between the mean annual precipitation primary driving force. in each study basin and streamﬂow (R) measured at the hydrological stations by neglecting inter-annual variabilities in water storage as: a Attribution of the temporal change in AED to climate change WBDE=− P R (8) The temporal changes in mean annual ETp, ETpa, and ETpr (ΔETp, The annual WBDE time series was restored to the level the reservoirs ΔETpa, and ΔETpr) were calculated between the two adjacent periods were at in 1961 (in order to exclude reservoir construction as the separated by the change point of annual ETp. These changes were fur- driving force of AC) by using the following formulas: ther quantitatively attributed to changes in the meteorological variables i NR using the sensitivity method of the Penman equation during the same ∑∑j==1962k 1 ()[(,)(,)]EEAjkAjkwLres−− riv WBDEii=+ WBDE period (Donohue et al., 2010), shown as follows: Absn ΔETpprp=+ ΔET ΔET a (12) (9)

A1 A1 WBDE− Ew r ∂ETpr dΔ ∂ETpr EL = ΔET dT dR (10) pr = ∫∫+ n 1 − r ∂Δ dT ∂Rn A0 A0 (13) 2015 ∑ KEpan, j j=1961 A1 ET ET ET Ew = ⎛ ∂ pa dΔ ∂ pa des ∂ pa ∂ea ⎞ 55 (11) ΔETpa = ∫ ⎜⎟+ + dT+ ∂Δ dT ∂es dT ∂ea ∂T A0 ⎝ ⎠ where WBDE is the WBDE for the ith year (mm); E and E are the i w L A1 A1 mean annual water and land surface evaporation in the study basins, ∂ETpa ∂ETpa ∂ea ∫∫du2+ dRH ∂u2 ∂ea ∂RH respectively (mm); Ares(j,k) is the reservoir area for jth year and kth A0 A0 (14) reservoir (km2), estimated by the mean value of the areas corresponding to the reservoir dead and normal storage levels; NR is the total where A refers to the variable sets that include T, u2, Rn, and RH. Corresponding subscripts 0 and 1 denote the states of A in the periods number of reservoirs for jth year; Ariv(j,k) are the original river areas (km2), estimated by the product of the length of backwater and the before and after the change point of AED, respectively. Δ Δ Δ Δ Δ width of the terminal of the backwater; A is the basin area (km2); The contribution rates of Rn, T, u2, and RH on ETp were es- bsn Δ WBDE is the mean annual WBDE (mm); r is the mean annual area timated by the ratio of the corresponding partial integral to ETp. The contribution rates on ΔET and ΔET were similarly estimated (Chu proportion of water bodies, estimated by the land use maps; Epan, j is the pr pa et al., 2017; Donohue et al., 2010; Ning, Li, Liu, & Han, 2016). Epan for jth year in the study basin; and K is a conversion coeﬃcient (Fu et al., 2004). b Attribution of the temporal change in AE to climate change and 3.2. Methods human activities

(1) Changes in annual evaporation-related meteorological variables The Budyko framework is a semi-empirical model used to quantify the water-energy balance under certain climate regions. Fu’s equation is Trends of annual evaporation-related meteorological variables were the analytical solution to the framework (Fu, 1981; Yang et al., 2007) ω detected by ordinary linear regression (OLR) (Zhang, Liu, Xu, Xu, & with parameter , describing the relationship between AED/P and AE/P (φ), and given as: Jiang, 2006) in terms of Rn, T, u2, RH, and P. One-sample ordinary Student’s t-test was applied to evaluate the trends (signiﬁcance 1 AED AED AED ω ω level = 0.05) (Gosset, 1942). φ = f ⎛ ,ω⎞ =+ 1 −⎡ 1 + ⎛ ⎞ ⎤ ⎝ P ⎠ P ⎣ ⎝ P ⎠ ⎦ (15) (2) The abnormal correlation between AED and AE Parameter ω is related to the geobotanic zonality types (Cai et al., 2014), which can be changed by human activities. Annual ω was cali-

AED usually shows a positive correlation with AE in Yangtze River brated based on observed φ (φo) and simulated φ (φs) according to OLS basin (Wang et al., 2011). In the study basins, the trends in annual ETp with relative error less than 5%, shown as follows: and Epan were detected by OLR and Student’s t-test to reveal any tem- 1 2 poral trends in AED (significance level = 0.05) (Gosset, 1942). Their ⎧ AED AED ω ω ⎫ minzφφ=− ( )2 = 1 +−⎡1 + ⎛ ⎞ ⎤ − φ change points were further assessed by the Pettitt method (Gao, Mu, so ⎨ P P o⎬ ⎣ ⎝ ⎠ ⎦ (16) Wang, & Li, 2011; Pettitt, 1979) at the same significance level. The ⎩ ⎭ trends in annual WBDE were similarly detected to reveal any temporal Calibrated annual ω was temporally averaged in the periods before trends in AE. Nonpositive correlations between AED and AE were the change point of AED (ω1) and after the point (ω2). Budyko curves fi identi ed if the trends of annual AED and AE were inconsistent from were drawn according to Fu’s equation, ω1 and ω2 in these two periods each other. Compared with the overall positive correlation in the (Fig. 2). Suppose the river basin has temporally shifted from Point S1 Yangtze River Basin, the detected nonpositive correlations in the study (prechange period) to Point S3 (postchange period). The φ and P in the basin were applied to verify the observed AC. prechange period was denoted as φ and P , which was changed into S1 S1 φ P S3 and S3 in the postchange period. The observed change in AE was (3) Driving force of the abnormal correlation between AED and AE assessed as (Wang & Hejazi, 2011):

ΔAEo =− φS PS3 φS PS1 (17) The driving force of AC was identiﬁed in three steps. Firstly, the 3 1 temporal change in AED was attributed to changes in meteorological The river basin would change from Point S1 to Point S2 along the variables according to the sensitivity method of the Penman equation curve obeying Fu’s equation, only driven by climate change (Fig. 2) (Donohue et al., 2010). Secondly, the observed temporal change in AE φ φ (Wang & Hejazi, 2011). The and P at the Point S2 were denoted as S2 was attributed to climate change and human activities according to the P ﬀ φ φ and S2, respectively. The di erence of the S2 and S1 was the climate decomposition method of the Budyko equation (Wang & Hejazi, 2011). change-induced change in φ, recorded as Δφc (Wang & Hejazi, 2011). Lastly, both attribution analyses were integrated to explore the impacts Then the climate change-induced change in AE (ΔAEc) was calculated

5 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Responses of temporal changes in AED and AE to climate change were revealed by combining the quantitative contribution rates of cli-

mate change on ΔETp with ΔAEc. In contrast, the response to human activities was detected by combining the inﬂuence of human activities

on ΔETp with ΔAEh. Both responses were integrated to determine why the AC appeared and then to identify its driving force. In addition, the annual area proportions of different land use types were compared to find the temporal-spatial distribution of land use during 1985–2010. The temporal changes in the annual FCR and NDVI were assessed to identify changes in land cover. The changes in land use and land cover were summarized to identify the main types of human activities. The detected main human activity was the primary driving force of AC, if the human activities were confirmed as the driving force.

4. Results

4.1. Changes in annual evaporation-related meteorological variables

4.1.1. A–P Model parameters at meteorological stations Table 2 shows the calibrated A–P Model parameters (a and b) and Fig. 2. Attributions of climate change and human activities to the change in evaluation at solar radiation stations. The parameters a and b ranged annual φ according to Fu’s equation. from 0.53 to 0.68 and 0.13 to 0.19, spatially averaged to be 0.57 and Note: φ is the ratio of actual evaporation to precipitation; P is precipitation; 0.16, respectively. The mean NSE value between the simulated and ω ω AED represents atmospheric evaporation demand; 1 and 2 represent the observed Rs was 0.85. And the Student’s t-test value was greater than its ’ parameters of Fu s equation before and after the change point of AED, respec- critical value (t0.05) at each station. Accordingly, the simulation of Rs by ω tively; S1 and S3 are the points characterized by the mean annual AED/P and the A–P model was considered to be acceptable. before and after the change point of AED, respectively; S2 is the point char- Table 3 shows the interpolated A–P Model parameters at meteor- acterized by the mean annual AED/P after the change point of AED and ω1. ological stations that did not have observed Rs using the Kriging method. The parameters a and b ranged from 0.53 to 0.66 and from as: 0.15 to 0.18, spatially averaged to be 0.58 and 0.16, respectively. ΔAE=− φ P φ P c S2 S2 S1 S1 (18) 4.1.2. Trends of annual evaporation-related meteorological variables The difference of the φ and φ was the human activities-induced S3 S2 Fig. 3 shows the temporal trends of basin-wide annual meteor- change in φ, recorded as Δφh (Wang & Hejazi, 2011). Then the human ological variables. Annual Rn, RH, and u2 decreased significantly in all activities-induced change in AE (ΔAE ) was calculated as: − h study basins at tendency rates ranging from -0.14 to -0.11 MJ m 2 day-1 ΔAE=− φ P φ P -1 − − -1 -1 -1 h S3 S3 S2 S2 (19) decade , 0.6 % to 0.4 % decade , and -0.16 to -0.10 m s decade , respectively. In contrast, annual T increased significantly at tendency -1 c Driving force of the abnormal correlation between AED and AE rates ranging from 0.17 to 0.28 °C decade depending on the different basins. Meanwhile, annual P exhibited no trend. The only exception to

Table 3

A-P Model parameters at meteorological stations that did not have observed Rs in southeastern China.

No. Name LON (°) LAT (°) abNo. Name LON (°) LAT (°) ab

1 Lianping 114.5 24.4 0.54 0.16 25 Yifeng 114.5 28.2 0.59 0.15 2 Longnan 114.5 24.6 0.54 0.15 26 Shangrao 117.6 28.3 0.59 0.18 3 Xunwu 115.7 25.0 0.53 0.17 27 Guixi 117.2 28.3 0.58 0.17 4 Nanxiong 114.3 25.1 0.56 0.15 28 Jingan 115.2 28.5 0.58 0.15 5 Shanghang 116.4 25.1 0.53 0.17 29 Dexing 117.4 28.6 0.59 0.17 6 Changting 116.4 25.9 0.54 0.17 30 Pingjiang 113.6 28.7 0.59 0.15 7 Guidong 113.6 26.1 0.59 0.15 31 Yushan 118.3 28.7 0.59 0.17 8 Ninghua 116.4 26.1 0.54 0.17 32 Xiushui 114.6 29.0 0.59 0.15 9 Ningdu 116.0 26.3 0.54 0.16 33 Poyang 116.7 29.0 0.61 0.15 10 Suichuan 114.5 26.3 0.57 0.15 34 Quzhou 118.9 29.0 0.59 0.16 11 Jiangganshan 114.2 26.6 0.58 0.15 35 Wuning 115.1 29.2 0.61 0.15 12 Guangchang 116.3 26.9 0.55 0.17 36 Jingdezhen 117.2 29.3 0.62 0.15 13 Taining 117.2 26.9 0.56 0.18 37 Gangkou 114.3 29.4 0.58 0.15 14 Lianhua 113.6 27.1 0.59 0.15 38 Honghu 113.3 29.5 0.55 0.15 15 Jianxian 114.9 27.1 0.56 0.15 39 Yangxin 115.1 29.5 0.62 0.15 16 Nanfeng 116.3 27.1 0.55 0.17 40 Qimen 117.4 29.5 0.62 0.15 17 Yongfeng 115.3 27.2 0.56 0.15 41 Chunan 119.0 29.6 0.59 0.15 18 Shaowu 117.5 27.3 0.57 0.18 42 Jiayu 113.9 30.0 0.54 0.16 19 Nancheng 116.7 27.6 0.56 0.17 43 Dongzhi 117.0 30.1 0.66 0.15 20 Yichun 114.4 27.8 0.59 0.15 44 Huangshan 118.2 30.1 0.59 0.15 21 Wuyishan 118.0 27.8 0.58 0.18 45 Huangshi 115.0 30.2 0.60 0.15 22 Zhuzhou 113.2 27.9 0.59 0.15 46 Jiangxia 114.2 30.2 0.54 0.16 23 Pucheng 118.5 27.9 0.59 0.18 47 Heta 116.2 30.3 0.66 0.15 24 Zhangshu 115.6 28.1 0.56 0.15

Note: LON denotes the longitude; LAT denotes the latitude; a and b denote the parameters of A–P model.

6 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Fig. 3. Annual variation of spatially averaged values of (a) temperature (T), (b) net radiation (Rn), (c) relative humidity (RH), (d) wind speed (u2), and (e) precipitation (P)infive river basins in southeastern China. (GRB, Ganjiang; FRB, Fuhe; XRB, Xinjiang; RRB Raohe; and XSB, Xiushui River basins). Note: solid lines denote significant temporal trends; dashed lines denote nonsignificant temporal trends.

7 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Fig. 4. Annual variation of spatially averaged values of potential evaporation (ETp), pan evaporation (Epan), and water budget-derived evaporation (WBDE) in river basins in southeastern China. Note: the panel letters (a, b, c, d, and e) represent Ganjiang (GRB), Fuhe (FRB), Xinjiang (XRB), Raohe (RRB), and Xiushui (XSB) River basins, respectively; solid lines denote signiﬁcant temporal trends; dashed lines denote nonsigniﬁcant temporal trends.

8 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Fig. 5. Change points of annual pan evaporation (Epan, a) and potential evaporation (ETp, b), in southeastern China. (GRB, Ganjiang; FRB, Fuhe; XRB, Xinjiang; RRB, Raohe; and XSB, Xiushui River basins).

these trends was that the Student’s t-test value for annual RH in XSB the decrease in ETpa. Accordingly, the influence of decreasing RH and was less than the critical value, meaning that no significant trend was increasing T was totally offset by the influence of the decreasing u2 to detected in that case. make annual ETpa constant, indicated by no trend detected (Supple- mentary Fig. 3). 4.2. The abnormal correlation between AED and AE Decreasing Rn and u2 contributed from 66 % to 110 % and 40%–70% to the decrease in ETp, respectively. Conversely, decreasing − − − Fig. 4 shows the temporal trends of basin-wide annual AED and AE. RH and increasing T contributed from 21 % to 6% and 37 % to −11 % to the decrease in ETp, respectively. Accordingly, decreasing Rn Annual ETp decreased significantly in all study basins except RRB at − − tendency rates ranging from −0.04 to −0.03 mm day 1 decade 1. and u2 produced larger variation in ETp.

Annual Epan decreased signiﬁcantly in all study basins except FRB at − − tendency rates ranging from −0.12 to −0.06 mm day 1 decade 1.In 4.3.2. Attribution of the temporal change in AE to climate change and contrast, annual WBDE increased in FRB at a tendency rate of 0.09 mm human activities − − day 1 decade 1 and remained constant in other study basins (Fig. 4). Fig. 7 shows the contributions of climate change and human ac-

The trends between AED and AE were different from each other. The tivities to the Δφo based on Fu’s equation. Mean annual equation observed spatial AC was verified. In addition, the slope value of the parameter ω increased from 2.02 before 1979 to 2.32 after 1979 in temporal trends of AED and AE tended to be smaller in GRB, FRB, and GRB, in addition to 2.52–3.23 in FRB, 1.68–1.72 in XRB, and 2.55–2.83 XRB after 1979 (see Supplementary Fig. 1 and 2). Temporal AC was also in RRB (Fig. 7a). In contrast, mean annual ω decreased slightly from verified in these basins. 1.82 to 1.76 in XSB. The Budyko curves were substantially higher after Fig. 5 shows the change points of annual AED. The maximum ab- 1979 than the Budyko curves before 1979 in GRB, FRB, XRB and RRB solute values of annual ETp’s Mann–Whitney statistics appeared in (Fig. 7b) except in XSB. 1979, 1979, 1979, and 1986 in GRB, FRB, XRB, and XSB respectively, Negative Δφc and ΔAEc values were detected, ranging from -0.057 to lying outside the confidence interval. These years were the change -0.025 and -22 to −17 mm respectively, depending on the different points of annual ETp. In comparison, the change points of annual Epan basins (Fig. 7c). The spatially averaged ΔAEc was −20 mm in the study were similarly determined to be 1992, 1979, 1979, and 1988 in GRB, basins. In contrast, positive Δφh and ΔAEh values were observed in four XRB, RRB, and XSB, respectively. The majority of the change points for of the five river basins, ranging from 0.01 to 0.05 and 40 to 141 mm both annual ETp and Epan occurred in 1979. respectively, depending on the different basins. Negative Δφh was only observed in XSB. The corresponding ΔAEh value was confirmed to be − Δ 4.3. Driving force of the abnormal correlation between AED and AE 5 mm. The spatially averaged AEh was 78 mm in the study basins. Accordingly, the overall climate change-induced decrease and the fi 4.3.1. Attribution of the temporal change in AED to climate change human activities-induced increase in AE was veri ed. Fig. 6 shows the contributions of the temporal changes in meteorological variables on the ΔETpr, ΔETpa, and ΔETp in the study basins. 4.3.3. Driving force of the abnormal correlation between AED and AE Increasing T contributed -33 % to -11 % to the decrease in ETpr (Fig. 6a) Decreasing Rn and u2 were the major climate factors reducing AED. by increasing the RLH (Supplementary Fig. 3). Conversely, the de- These reductions also led to a 20 mm decrease in mean annual AE in the creasing Rn contributed 111 %–133 % to the decrease in ETpr. Ac- study basins. Climate change led to the simultaneous temporal decrease cordingly, an increased proportion of latent heat relative to Rn was in AED and AE. Human activities resulted in an increase of 78 mm in insufficient to compensate for the decrease in ETpr induced by de- mean annual AE but had no effect on AED. Eventually, annual AED creasing Rn, and as such ETpr decreased. decreased, but annual AE did not appear to exhibit the same temporal Decreasing RH and increasing T contributed −126 % to −18 % and trend. So overall, human activities appear to be the driving force of AC.

−59 % to −11 % to the decrease in ETpa in the study basins, respec- Fig. 8 shows the temporal-spatial distribution of land use and cover. tively (Fig. 6b). Conversely, deceasing u2 contributed 128 %–284 % to The forest land and arable land accounted for 62.7 % and 26.1 % of the

9 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Fig. 6. Contribution rates of the temporal changes in meteorological variables to changes in (a) radiation fraction of Penman potential evaporation (ETpr), (b) aerodynamic fraction of Penman potential evaporation (ETpa) and (c) Penman potential evaporation (ETp) in river basins in southeastern China.

Note: GRB, FRB, XRB, RRB, and XSB are the Ganjiang, Fuhe, Xinjiang, Raohe, and Xiushui River basins, respectively; ΔRn, ΔT, ΔRH, and Δu2 denote the temporal changes in net radiation, temperature, relative humidity, and wind speed respectively. terrestrial surface, respectively. These two land use types remained and afforestation to the temporal change in AE. constant during the period from 1985 to 2010, indicated by relatively small standard errors. Accordingly, the forest and arable land were the main land use types. Additionally, both of the annual FCR and NDVI 5.1. Changes in annual evaporation-related meteorological variables simultaneously and dramatically increased around 1980. Arable land consisted mainly of double-cropping paddy fields, whose NDVI fluc- Meteorological states are regulated by both local water-energy tuated during half-year periods rather than multi-year periods. There- processes and large-scale climate conditions. In the five river basins fore, the growth of paddy areas could not cause the observed temporal used in this study, the large-scale climate conditions were especially increase in annual NDVI. So, the observed increase in NDVI and FCR important. Specifically, decreasing Rn and u2 are the likely consequence was a result of afforestation. Accordingly, afforestation was the main of radiative dimming (Cohen, Ianetz, & Stanhill, 2002; Liu et al., 2004; human activity leading to the abnormal correlation between AED and Papaioannou, Kitsara, & Athanasatos, 2011; Roderick & Farquhar, AE rather than urbanization and climate change. 2002; Wild, Ohmura, Gilgen, & Rosenfeld, 2004) and wind stilling (McVicar, Roderick, Donohue, Li et al., 2012; Roderick, Rotstayn, Farquhar, & Hobbins, 2007), respectively. Increasing annual tempera- 5. Discussion ture was consistent with global warming that is induced by increasing concentrations of greenhouse gases in the atmosphere (Milly & Dunne, This study focused on identifying and interpreting the AC between 2016; Tsai, 2017; Vallis, Zurita-Gotor, Cairns, & Kidston, 2015). Annual AED and AE in the energy-limited region of southeastern China. precipitation showed no trend in the study basins, similar to overall Afforestation was determined to be the major human activity that likely trends in the Yangtze River basin (Ukkola & Prentice, 2013). explains the observed AC. During this analysis process, we quantitatively established the integrated response chain from climate change

10 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Fig. 7. Quantitative attributions on Δφo based on Fu’s equation. Note: the solid lines and dashed lines in each panel denote the periods before and after 1979, respectively; GRB, FRB, XRB, RRB, and XSB are the Ganjiang, Fuhe, Xinjiang, Raohe, and Xiushui River basins, respectively; ω represents the parameter of Fu’s equation; P is precipitation; WBDE is water-budget derived evaporation;

ETp is potential evaporation; Δφo, Δφc, and Δφh are the observed, climate change-induced and human activities-induced ratio of actual evaporation to precipitation, respectively; S1 and S3 are the points characterized by the mean annual ETp/P and ω before and after the change point of AED, respectively; S2 is the point characterized by the mean annual AED/P after the change point of AED and ω before the point.

Fig. 8. Spatial framework and temporal changes of land use and cover. FCR, forest cover rate; NDVI, normalized diﬀerence vegetation index. The error bars are the standard errors of the means.

11 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

5.2. The abnormal correlation between AED and AE et al., 2013). Consequently, understanding LUCC may sometimes be uncertain. There may also be unknown genetic effects on changes in Study basins are an intermediate-scale region. At such a scale, a water-energy flows. Combining land use maps with vegetation canopy stable correlation between AED and AE can be expected by eliminating change indices may be a better way to understand LUCC (Ye, Zhang, the impact of the micro-topography. The driving forces of AC are re- Bai, & Hu, 2011). stricted to climate change and human activities rather than the spatial distribution of the micro-topography. Specifically, at smaller spatial 6. Conclusion scales, diverse correlations between AED and AE may be detected. This diversity may be regulated more by the complicated spatial distribution We assessed the temporal changes in the meteorological variables, of micro-topography. However, as the spatial scale under consideration AED, and AE by ordinary linear regression and by the Pettitt method for is expanded, such diversities are possibly masked such that a single type five energy-limited basins in southeast China. The temporal trends be- of correlation between AED and AE is observed. With scale expansion, tween AED and AE were compared to explore whether the abnormal the influence of microtopography on the correlation gradually becomes correlations appeared in the study basins relative to the Yangtze River weaker. Under this condition, a stable correlation represents the in- basin. The driving forces of the abnormal correlation were identified as teracting effects of local climate and terrestrial land surface condition being either climate change or human activities using the sensitivity on the hypothesis of no effect of human activities. The AC can be at- method of the Penman equation and the decomposition method of the tributed to climate and human activities-induced terrestrial land sur- Budyko equation. face changes. Annual Rn, u2, and RH were found to decrease significantly at ten- −2 -1 -1 AED decreased in the study basins, as indicated by ETp and Epan. This dency rates ranging from -0.14 to -0.11 MJ m day decade , -0.6 % finding was consistent with overall temporally decreasing trends of AED to -0.4 % decade-1, and -0.16 to -0.10 m s-1 decade-1, respectively. in the world mentioned above. However, a significant change point for Conversely, annual T increased significantly at tendency rates ranging AED occurred in 1979 when AED began to decrease. After 1979, the from 0.17 to 0.28 °C decade-1. Meanwhile, annual P was found to have study basins also experienced strong afforestation. This was suspected no trend. as the driving force causing the existence of the change point. Generally, AED showed a positive correlation with AE in the

AC between AED and AE was confirmed for the study basins pre- Yangtze River basin. Annual ETp and Epan decreased significantly at − sented in this paper. In humid regions, the observation and simulation tendency rates ranging from -0.04 to -0.03 and -0.12 to -0.06 mm d 1 − on the AC have been reported for areas with dense forests (Carmona decade 1, respectively. Their change points both occurred in 1979. et al., 2016; Lawrimore & Peterson, 2000). Perhaps the areas of these Accordingly, annual AED decreased after 1979. In comparison, AE in- − − previous studies experienced similar afforestation as occurred in our creased significantly at a tendency rate of 0.09 mm d 1 decade 1 in study basins. This supposition needs to be further verified. FRB and remained constant in other study basins, as indicated by WBDE. Nonpositive correlation between AED and AE appeared and AC 5.3. Driving force of the abnormal correlation between AED and AE was verified compared with the Yangtze River basin. The increasing T contributed −33 % to −11 % to the decrease in

Local hydrological processes and large-scale climate conditions are ETpr by increasing the proportion of latent heat to Rn in the study ba- the key variables regulating negative and positive correlation between sins. This increased proportion of latent heat cannot compensate for the

AED and AE, respectively. Specifically, an arid region was typically far decrease in ETpr induced by a reduction in Rn, the contribution rates of away from oceans. This region can be likened to a relatively enclosed which ranged from 111 %–133 %. Eventually, annual ETpr decreased room with little outside influence. The changes in water-energy flows significantly at tendency rates ranging from −0.03 to −0.018 mm − − during the local hydrological process, indicated by the changes in AE, day 1 decade 1. In addition, decreasing RH and increasing T con- may effectively alter the local climate. During this process, the AED tributed −126 % to −18 % and −59 % to −11 %, respectively, to the changed through inverse trends of AE, which can be understood as a decrease in ETpa by increasing VPD. This influence was totally offset by compensation mechanism for AE changes according to BCH (Bouchet, of the influence of decreasing u2, the contribution rates of which ranged 1963). This compensation mechanism was more significant in arid re- from 128 %–284 %. There was, consequently, no temporal trend de- gions because of strong local interaction. Consequently, the negative tected in ETpa. Meanwhile, decreasing Rn and u2 contributed most to the correlation between AED and AE was easily seen in this region. In decrease in ETp, with contribution rates ranging from 66 % to 110 % contrast, the humid region was closer to the ocean. The correlation was and 40 %–70 %, respectively. Accordingly, decreasing Rn and u2 were regulated by large-scale climate conditions. The impact of local water- the major climate factors in reducing AED. energy processes on the correlation between AED and AE was relatively Climate change also led to the simultaneous decrease of 20 mm in limited and easily offset by large-scale climate conditions. For this si- mean annual AE in the study basins. However, human activities had no tuation, the positive correlation was easily observed in this region. effect on AED and resulted in an increase of 78 mm in mean annual AE. Afforestation strengthened the local hydrological process, thus Eventually annual AED decreased, but annual AE did not exhibit the driving the correlation between AED and AE to tend to be negative. same temporal trend. Accordingly, afforestation was the main human Specifically, the study basins were typically energy-limited regions, activity changing the correlation between AED and AE (rather than where the decrease in AE was expected as a result of decreasing AED. urbanization and climate change). However, the regional afforestation increased the ω parameter of Fu’s equation and the ratio of AE to AED. This afforestation-induced ele- Acknowledgement vated ratio offset the expected decrease in AE against the impact of decreasing AED. It produced the inconsistent temporal trends between The contribution of Xingmin Mu and Guangju Zhao presents the AED and AE through converting more atmospheric evaporation energy proposed hypothesis and design of the research framework. Hua Bai are into actual one, which strengthened the local hydrological process. As a responsible for the data analysis, followed by writing paper, sponsored result, the correlation between AED and AE tended to be negative. by Science and Technology Project for the Education Department of Accurate assessment of land use and coverage changes (LUCC) re- Jiangxi Province (GJJ171005) and the key program for the Jiangxi quires integration of changes in land use maps and vegetation canopy Provincial Hydraulic Science and Technology Projects (KT201726). data. However, current simulations of streamflow or evaporation typi- Xiaoxiao Yang and Jianchu Huang are responsible for the data collec- cally consider only either land use maps or a vegetation canopy change tion and quality control. This paper is revised by Xianghui Lu, Chao Yue index, such as NDVI or leaf area index (Guo, Hu, & Jiang, 2008; Yan and Faliang Gui.

12 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

Appendix A. Supplementary data adaptation to mega-drought: Anticipatory water modeling, policy, and planning for the urban Southwest. Sustainable Cities and Society, 27, 497–504. https://doi.org/10. 1016/j.scs.2016.05.001. Supplementary material related to this article can be found, in the Golubev, V. S., Lawrimore, J. H., Groisman, P. Y., Speranskaya, N. A., Zhuravin, S. A., online version, at doi:https://doi.org/10.1016/j.scs.2020.102075. Menne, M. J., et al. (2001). Evaporation changes over the contiguous United States and the former USSR: A reassessment. Geophysical Research Letters, 28(13), 2665–2668. https://doi.org/10.1029/2000GL012851. References Gosset, W. S. (1942). In E. S. Pearson, & J. Wishart (Eds.). " Student’s" collected papers. Cambridge: Cambridge university press. Abtew, W., Obeysekera, J., & Iricanin, N. (2010). Pan evaporation and potential evapo- Guo, H., Hu, Q., & Jiang, T. (2008). Annual and seasonal streamflow responses to climate transpiration trends in South Florida. Hydrological Processes, 25(6), 958–969. https:// and land-cover changes in the Poyang Lake basin. China. Journal of Hydrology, 355(1- doi.org/10.1002/hyp.7887. 4), 106–122. https://doi.org/10.1016/j.jhydrol.2008.03.020. Ali, A. M., Shafiee, M. E., & Berglund, E. Z. (2017). Agent-based modeling to simulate the Hardin, E., AghaKouchak, A., Qomi, M. J. A., Madani, K., Tarroja, B., Zhou, Y., et al. dynamics of urban water supply: Climate, population growth, and water shortages. (2017). California drought increases CO2 footprint of energy. Sustainable Cities and Sustainable Cities and Society, 28, 420–434. https://doi.org/10.1016/j.scs.2016.10. Society, 28, 450–452. https://doi.org/10.1016/j.scs.2016.09.004. 001. Hember, R. A., Coops, N. C., & Spittlehouse, D. L. (2017). Spatial and temporal variability Allen, R. G., Pereira, L. S., Raes, D., & Smith, M. (1998). Crop evapotranspiration-Guidelines of potential evaporation across North American forests. Hydrology, 4,5.https://doi. for computing crop water requirements. FAO Irrigation and Drainage Paper, 56Rome: org/10.3390/hydrology4010005. FAO. Hobbins, M. T., Dai, A., Roderick, M. L., & Farquhar, G. D. (2008). Revisiting the para- Angstrom, A. (1924). Solar and terrestrial radiation. Report to the international com- meterization of potential evaporation as a driver of long-term water balance trends. mission for solar research on actinometric investigations of solar and atmospheric Geophysical Research Letters, 35, L12403. https://doi.org/10.1029/2008GL033840. radiation. Quarterly Journal of the Royal Meteorological Society, 50(210), 121–126. Hong, J., Zhong, X., Guo, S., Liu, G., Shen, G. Q., & Yu, T. (2019). Water-energy nexus and https://doi.org/10.1002/qj.49705021008. its efficiency in China’s construction industry: Evidence from province-level data. Arora, V. (2002). Modeling vegetation as a dynamic component in soil-vegetation-at- Sustainable Cities and Society, 48, 101557. https://doi.org/10.1016/j.scs.2019. mosphere transfer schemes and hydrological models. Reviews of Geophysics, 40(2), 3- 101557. 1-3-26. https://doi.org/10.1029/2001RG000103. Lawrimore, J. H., & Peterson, T. C. (2000). Pan evaporation trends in dry and humid Bouchet, R. J. (1963). Evapotranspiration réelle et potentielle, signification climatique. regions of the United States. Journal of Hydrometeorology, 1, 543–546. https://doi. IAHS Publications, 62, 134–142. org/10.1175/1525-7541(2000)001<0543:PETIDA>2.0.CO;2. Brutsaert, W., & Parlange, M. B. (1998). Hydrologic cycle explains the evaporation Liu, B., Xu, M., Henderson, M., & Gong, W. (2004). A spatial analysis of pan evaporation paradox. Nature, 396, 30. https://doi.org/10.1038/23845. trend in China, 1955-2000. Journal of Geophysical Research Atmospheres, 109, Budyko, M. I. (Ed.). (1974). Climate and life, translated from Russian by D. H. Miller. New D15102. https://doi.org/10.1029/2004JD004511. York: Elsevier. McVicar, T. R., Roderick, M. L., Donohue, R. J., Li, L. T., Niel, T. G. V., Thomas, A., et al. Bureau of Statistics of Jiangxi (2012). Jiangxi statistical yearbook. Beijing: China Statistics (2012). Global review and synthesis of trends in observed terrestrial near-surface Press (in Chinese). wind speeds: Implications for evaporation. Journal of Hydrology, 416-417, 182–205. Cai, D., Fraedrich, K., Sielmann, F., Guan, Y., Guo, S., Zhang, L., et al. (2014). Climate and https://doi.org/10.1016/j.jhydrol.2011.10.024. vegetation: An ERA-Interim and GIMMS NDVI analysis. Journal of Climate, 27(13), McVicar, T. R., Roderick, M. L., Donohue, R. J., & Niel, T. G. V. (2012). Less bluster 5111–5118. https://doi.org/10.1175/JCLI-D-13-00674.1. ahead? Ecohydrological implications of global trends of terrestrial near-surface wind Carmona, A. M., Poveda, G., Sivapalan, M., Vallejo-Bernal, S. M., & Bustamante, E. speeds. Ecohydrology, 5(4), 381–388 http://doi.org/10.1002/eco.1298. (2016). A scaling approach to Budyko’s framework and the complementary re- Milly, P. C. D. (1994). Climate, soil water storage, and the average annual water balance. lationship of evapotranspiration in humid environments: Case study of the Amazon Water Resources Research, 30(7), 2143–2156. https://doi.org/10.1029/94WR00586. River basin. Hydrology and Earth System Sciences, 20, 589–603. https://doi.org/10. Milly, P. C. D., & Dunne, K. A. (2016). Potential evapotranspiration and continental 5194/hess-20-589-2016. drying. Nature Climate Change, 6, 946–949. https://doi.org/10.1038/ Chattopadhyay, N., & Hulme, M. (1997). Evaporation and potential evapotranspiration in NCLIMATE3046. India under conditions of recent and future climate change. Agricultural and Forest Moene, A. F., Beyrich, F., & Hartogensis, O. K. (2009). Developments in scintillometry. Meteorology, 87(1), 55–73. https://doi.org/10.1016/S0168-1923(97)00006-3. Bulletin of the American Meteorological Society, 90(5), 694–698. https://doi.org/10. Chen, D., Gao, G., Xu, C., Guo, J., & Ren, G. (2005). Comparison of the Thornthwaite 1175/2008BAMS2672.1. method and pan data with the standard Penman-Monteith estimates of reference Moonen, A. C., Ercoli, L., Mariotti, M., & Masoni, A. (2002). Climate change in Italy evapotranspiration in China. Climate Research, 28(2), 123–132. https://doi.org/10. indicated by agromete-orological indices over 122 years. Agricultural and Forest 3354/cr028123. Meteorology, 111(1), 13–27. https://doi.org/10.1016/S0168-1923(02)00012-6. Chu, R., Li, M., Shen, S., Islam, A. R. M. T., Cao, W., Tao, S., et al. (2017). Changes in Nachshon, U., Netzer, L., & Livshitz, Y. (2016). Land cover properties and rain water reference evapotranspiration and its contributing factors in Jiangsu, a major eco- harvesting in urban environments. Sustainable Cities and Society, 27, 398–406. nomic and agricultural province of eastern China. Water, 9(7), 486. https://doi.org/ https://doi.org/10.1016/j.scs.2016.08.008. 10.3390/w9070486. Nakamichi, T., & Moroizumi, T. (2015). Applicability of three complementary relation- Cohen, S., Ianetz, A., & Stanhill, G. (2002). Evaporative climate changes at Bet Dagan, ship models for estimating actual evapotranspiration in urban area. Journal of Israel, 1964–1998. Agricultural and Forest Meteorology, 111(2), 83–91. https://doi. Hydrology and Hydromechanics, 63(2), 117–123. https://doi.org/10.1515/johh-2015- org/10.1016/S0168-1923(02)00016-3. 0011. Computer Network Information Center (2009). Chinese academy of sciences. 30 m resolu- Nash, J. E., & Sutcliffe, J. V. (1970). River flow forecasting through conceptual models tion GDEM data. DEM data. Geospatial data cloud site. (in Chinese) http://www. part I-A discussion of principles. Journal of Hydrology, 10, 282–290. https://doi.org/ gscloud.cn/. 10.1016/0022-1694(70)90255-6. Croitoru, A., Piticar, A., Dragotă, C. S., & Burada, D. C. (2013). Recent changes in re- National Meteorological Information Center (2019). Dataset of daily climate data from ference evapotranspiration in Romania. Global and Planetary Change, 111, 127–136. Chinese surface stations. Surface data and products, v3.0. (in Chinese) http://data.cma. https://doi.org/10.1016/j.gloplacha.2013.09.004. cn/site/index.html. DeBruin, H. A. R. (2009). Time to think: Reflections of a pre-pensioned scintillometer Ning, T., Li, Z., Liu, W., & Han, X. (2016). Evolution of potential evapotranspiration in the researcher. Bulletin of the American Meteorological Society, 90(5), ES17–ES26. https:// northern Loess Plateau of China: recent trends and climatic drivers. International doi.org/10.1175/2008BAMS2704.2. Journal of Climatology, 36(12), 4019–4028. https://doi.org/10.1002/joc.4611. Donohue, R. J., McVicar, T. R., & Roderick, M. L. (2010). Assessing the ability of potential Ozturk, M. (2015). An evaluation of global solar radiation empirical formulations in evaporation formulations to capture the dynamics in evaporative demand within a Isparta. Turkey. Energy Sources, 37(22), 2474–2486. https://doi.org/10.1080/ changing climate. Journal of Hydrology, 386(1-4), 186–197. http://10.1016/j.jhydrol. 15567036.2012.745037. 2010.03.020. Papaioannou, G., Kitsara, G., & Athanasatos, S. (2011). Impact of global dimming and Fu, B. (1981). On the calculation of the evaporation from land surface. Chinese Journal of brightening on reference evapotranspiration in Greece. Journal of Geophysical Atmospheric Sciences, 5(1), 23–31 (in Chinese). Research, 116, D09107. https://doi.org/10.1029/2010JD015525. Fu, G., Liu, C., Chen, S., & Hong, J. (2004). Investigating the conversion coefficients for Penman, H. L. (1948). Natural evaporation from open water, bare soil and grass. free water surface evaporation of different evaporation pans. Hydrological Processes, Proceedings of the Royal Society of London Series A, Mathematical and Physical Sciences, 18(12), 2247–2262. https://doi.org/10.1002/hyp.5526. 193(1032), 120–145. Fu, G., Charles, S. P., & Yu, J. (2009). A critical overview of pan evaporation trends over Peterson, T. C., Golubev, V. S., & Groisman, P. Y. (1995). Evaporation losing its strength. the last 50 years. Climatic Change, 97(1), 193–214. https://doi.org/10.1007/s10584- Nature, 377, 687–688. https://doi.org/10.1038/377687b0. 009-9579-1. Pettitt, A. N. (1979). A non-parametric approach to the change-point problem. Journal of Gao, P., Mu, X., Wang, F., & Li, R. (2011). Changes in streamflow and sediment discharge the Royal Statistical Society Series C, Applied Statistics, 28(2), 126–135. http://10. and the response to human activities in the middle reaches of the Yellow River. 2307/2346729. Hydrology and Earth System Sciences Discussions, 7, 6793–6822. https://doi.org/10. Pinzon, J. E., & Tucker, C. J. (2016). GIMMS NDVI3g Data. Ecocast data, v1. https:// 5194/hessd-7-6793-2010. ecocast.arc.nasa.gov/data/pub/gimms/3g.v1/. Giridharan, R., & Emmanuel, R. (2018). The impact of urban compactness, comfort Pinzon, J. E., & Tucker, C. J. (2014). A non-stationary 1981–2012 AVHRR NDVI3g time strategies and energy consumption on tropical urban heat island intensity: A review. series. Remote Sensing, 6(8), 6929–6960. https://doi.org/10.3390/rs6086929. Sustainable Cities and Society, 40, 677–687. https://doi.org/10.1016/j.scs.2018.01. Remondi, F., Burlando, P., & Vollmer, D. (2016). Exploring the hydrological impact of 024. increasing urbanisation on a tropical river catchment of the metropolitan Jakarta. Gober, P., Sampson, D. A., Quay, R., White, D. D., & Chow, W. T. L. (2016). Urban Indonesia. Sustainable Cities and Society, 20, 210–221. https://doi.org/10.1016/j.scs.

13 H. Bai, et al. Sustainable Cities and Society 56 (2020) 102075

2015.10.001. Society, 51, 101722. https://doi.org/10.1016/j.scs.2019.101722. Roads, J. (2003). The NCEP-NCAR, NCEP-DOE, and TRMM tropical atmosphere hydro- Wang, Y., Liu, B., Zhai, J., Su, B., Luo, Y., & Zhang, Z. (2011). Relationship between logic cycles. Journal of Hydrometeorology, 4(5), 826–840. https://doi.org/10.1175/ potential and actual evaporation in Yangtze River basin. Advances in Climate Change 1525-7541(2003)004<0826:TNNATT>2.0.CO;2. Research, 7(6), 393–399 (In Chinese with English abstract). Roberge, F., & Sushama, L. (2018). Urban heat island in current and future climates for Wang, Z., Xie, P., Lai, C., Chen, X., Wu, X., Zeng, Z., et al. (2017). Spatiotemporal the island of Montreal. Sustainable Cities and Society, 40, 501–512. https://doi.org/10. variability of reference evapotranspiration and contributing climatic factors in China 1016/j.scs.2018.04.033. during 1961–2013. Journal of Hydrology, 544,97–108. https://doi.org/10.1016/j. Roderick, M. L., & Farquhar, G. D. (2002). The cause of decreased pan evaporation over jhydrol.2016.11.021. the past 50 years. Science, 298(5597), 1410–1411. http://10.1126/science. Wild, M., Ohmura, A., Gilgen, H., & Rosenfeld, D. (2004). On the consistency of trends in 1075390-a. radiation and temperature records and implications for the global hydrological cycle. Roderick, M. L., & Farquhar, G. D. (2004). Changes in Australian pan evaporation from Geophysical Research Letters, 31, L11201. https://doi.org/10.1029/2003GL019188. 1970 to 2002. International Journal of Climatology, 24(9), 1077–1090. https://doi. Wolock, D. M., & McCabe, G. J. (1999). Explaining spatial variability in mean annual org/10.1002/joc.1061. runoff in the conterminous United States. Climate Research, 11, 149–159. https://doi. Roderick, M. L., & Farquhar, G. D. (2005). Changes in New Zealand pan evaporation since org/10.3354/cr011149. the 1970s. International Journal of Climatology, 25(15), 2031–2039. https://doi.org/ World Meteorological Organization (WMO) (2014). Measurement of evaporation, in WMO 10.1002/joc.1262. guide to meteorological instruments and methods of observationSwitzerland: Geneva, Roderick, M. L., Rotstayn, L. D., Farquhar, G. D., & Hobbins, M. T. (2007). On the at- (Chapter 10). Available at(8th ed.). https://library.wmo.int/opac/doc_num.php? tribution of changing pan evaporation. Geophysical Research Letters, 34(17), L17403 explnum_id=3156. http://doi.org/1029/2007GL031166. Xu, Y., Pan, S., Gao, C., Fu, G., & Chiang, Y. (2016). Historical pan evaporation changes in Sarangi, A., Madramootoo, C. A., & Enright, P. (2006). Comparison of spatial variability the Qiantang River Basin, East China. International Journal of Climatology, 36, techniques for runoff estimation from a Canadian watershed. Biosystems Engineering, 1928–1942. https://doi.org/10.1002/joc.4470. 95(2), 295–308. https://doi.org/10.1016/j.biosystemseng.2006.06.002. Yan, D., Schneider, U. A., Schmid, E., Huang, H. Q., Pan, L., & Dilly, O. (2013). Shen, Y., Liu, C., Liu, M., Zeng, Y., & Tian, C. (2010). Change in pan evaporation over the Interactions between land use change, regional development, and climate change in past 50 years in the arid region of China. Hydrological Processes, 24(2), 225–231. the Poyang Lake district from 1985 to 2035. Agricultural Systems, 119,10–21. https:// https://doi.org/10.1002/hyp.7435. doi.org/10.1016/j.agsy.2013.04.001. Shuttleworth, W. J. (1993). Evaporation in handbook of hydrology. Sydney: McGraw-Hill. Yang, H., & Yang, D. (2012). Climatic factors influencing changing pan evaporation Solignac, P. A., Brut, A., Selves, J. L., Béteille, J. P., Gastellu-Etchegorry, J. P., Keravec, P., across China from 1961 to 2001. Journal of Hydrology, 414-415, 184–193. https:// et al. (2009). Uncertainty analysis of computational methods for deriving sensible doi.org/10.1016/j.jhydrol.2011.10.043. heat flux values from scintillometer measurements. Atmospheric Measurement Yang, D., Sun, F., Liu, Z., Cong, Z., & Lei, Z. (2006). Interpreting the complementary Techniques, 2, 1383–1417. https://doi.org/10.5194/amt-2-741-2009. relationship in non-humid environments based on the Budyko and Penman hy- Stanhill, G., & Möller, M. (2008). Evaporative climate change in the British Isles. potheses. Geophysical Research Letters, 33(18), L18402. https://doi.org/10.1029/ International Journal of Climatology, 28(9), 1127–1137. https://doi.org/10.1002/joc. 2006GL027657. 1619. Yang, D., Sun, F., Liu, Z., Cong, Z., Ni, G., & Lei, Z. (2007). Analyzing spatial and temporal Sun, F. (2007). Study on watershed evapotranspiration based on the Budyko Hypothesis. variability of annual water-energy balance in nonhumid regions of China using the Beijing: Tsinghua University 114 pp. (In Chinese with English abstract). Budyko hypothesis. Water Resources Research, 43, W04426. https://doi.org/10.1029/ Thomas, A. (2000). Spatial and temporal characteristics of potential evapotranspiration 2006WR005224. trends over China. International Journal of Climatology, 20(4), 381–396. https://doi. Ye, X., Zhang, Q., Bai, L., & Hu, Q. (2011). A modeling study of catchment discharge to org/10.1002/(SICI)1097-0088(20000330)20:4<381::AID-JOC477>3.0.CO;2-K. Poyang Lake under future climate in China. Quaternary International, 244(2), Tsai, W. (2017). Implementation status of Taiwan carbon footprint (CFP) system and a 221–229. https://doi.org/10.1016/j.quaint.2010.07.004. case study of the Taiwan High Speed Rail Corporation. Sustainable Cities and Society, Zhang, L., Traore, S., Cui, Y., Luo, Y., Zhu, G., Liu, B., et al. (2019). Assessment of spa- 35, 331–335. https://doi.org/10.1016/j.scs.2017.08.021. tiotemporal variability of reference evapotranspiration and controlling climate fac- Ukkola, A. M., & Prentice, I. C. (2013). A worldwide analysis of trends in water-balance tors over decades in China using geospatial techniques. Agricultural Water evapotranspiration. Hydrology and Earth System Sciences, 17(10), 4177–4187. https:// Management, 213(C), 499–511. https://doi.org/10.1016/j.agwat.2018.09.037. doi.org/10.5194/hess-17-4177-2013. Zhang, Q., Liu, C., Xu, C., Xu, Y., & Jiang, T. (2006). Observed trends of annual maximum Vallis, G. K., Zurita-Gotor, P., Cairns, C., & Kidston, J. (2015). Response of the large-scale water level and streamflow during past 130 years in the Yangtze River basin. China. structure of the atmosphere to global warming. Quarterly Journal of the Royal Journal of Hydrology, 324(1-4), 255–265. https://doi.org/10.1016/j.jhydrol.2005.09. Meteorological Society, 141(690), 1479–1501. https://doi.org/10.1002/qj.2456. 023. Wang, K., & Dickinson, R. E. (2012). A review of global terrestrial evapotranspiration: Zhang, Y., Leuning, R., Chiew, F. H. S., Wang, E., Zhang, L., Liu, C., et al. (2012). Decadal Observation, modeling, climatology, and climatic variability. Reviews of Geophysics, trends in evaporation from global energy and water balances. Journal of 50(2), RG2005. https://doi.org/10.1029/2011RG000373. Hydrometeorology, 13(1), 379–391. https://doi.org/10.1175/JHM-D-11-012.1. Wang, D., & Hejazi, M. (2011). Quantifying the relative contribution of the climate and Zuo, H., Chen, B., Wang, S., Guo, Y., Zuo, B., Wu, L., et al. (2016). Observational study on direct human impacts on mean annual streamflow in the contiguous United States. complementary relationship between pan evaporation and actual evapotranspiration Water Resources Research, 47, W00J12. https://doi.org/10.1029/2010WR010283. and its variation with pan type. Agricultural and Forest Meteorology, 222,1–9. https:// Wang, Y., Feng, Y., Zuo, J., & Rameezdeen, R. (2019). From “Traditional” to “Low doi.org/10.1016/j.agrformet.2016.03.002. carbon” urban land use: Evaluation and obstacle analysis. Sustainable Cities and